The goal of this paper is to evaluate the shear capacity of fibrous recycled aggregate concrete beams side bonded with carbon fiber-reinforced polymer (CFRP) based on numerical modeling and deep learning. Forty-five models were numerically investigated in ANSYS with different lengths and layers of vertical CFRP sheets. Concrete properties built for coarse recycled aggregate and various fibers as 3D and 5D steel, synthetic, and hybrid fibers added at a 0.75% volume fraction. Finally, feed forward, cascade-forward back propagation, and NARX network algorithms were adopted to predict the shear capacity. Shear capacity improvement was significantly affected by CFRP rehabilitation. Double-layered rehabilitation applied on the zone between support and load was the most adequate. The performance of cascade-forward back propagation in the estimation of shear capacity was superior to the other deep auto encoder models. The present paper offers good guidance on the practice field as a valuable supplement to the laboratory investigation and theoretical calculations.